Computer system and method for automated batch data alignment in batch process modeling, monitoring and control

Abstract

Embodiments include a computer-implemented method (and system) for performing automated batch data alignment for modeling, monitoring, and control of an industrial batch process. The method (and system) loads, scales, and screens plant historian batch data for an industrial batch process. The method (and system) selects a reference batch as basis of the batch alignment, defines and adds or modifies one or more batch phases, and selects one or more batch variables based on one or more profiles and corresponding curvatures of the batch data. The method (and system) estimates one or more weightings, adjust one or more tuning parameters and uses a sliding time window combined with DTW, DTI and GSS algorithms, performs the batch alignment in offline mode or online mode.

What is claimed is: 
     
       1. A computer-implemented method for performing automated batch data alignment for modeling, monitoring, and controlling an industrial batch process, the method comprising:
 loading batch data from a plant historian database for a subject industrial batch process; 
 scaling the loaded batch data for batch alignment; 
 screening and removing outliers of batch data from the scaled batch data; 
 selecting a reference batch as basis of the batch alignment from the screened batch data; 
 defining and adding or modifying one or more batch phases associated with the screened batch data for the batch alignment; 
 selecting one or more batch variables associated with the screened batch data based on at least one of: (i) one or more profiles of the one or more batch variables and (ii) one or more curvatures of the one or more batch variables; 
 estimating one or more variable weightings based upon the selected batch variables associated with the batch alignment; 
 selecting a given batch of the screened batch data to perform the batch alignment against the reference batch; 
 adjusting one or more tuning parameters associated with the batch alignment; and 
 performing the batch alignment in at least one of an offline mode and an online mode, wherein performing the batch alignment includes performing alignment for a sliding window associated with the subject industrial batch process based upon a continuous objective function that includes a linearly interpolated value solved using dynamic time interpolation (DTI). 
 
     
     
       2. The method of  claim 1 , wherein the screening and removing the outliers of the batch data comprises one or more of the steps of:
 a) further screening measurements of the selected batch variables for irregular behaviors as compared to behaviors associated with other batches of the screened batch data; 
 b) repairing data of the screened batch data associated with the irregular behaviors; 
 c) removing one or more invalid batches from the screened batch data after the data repair including batches that include different variation profiles as compared to the other batches over a time interval; 
 d) resampling the selected batch variables with a base sampling rate; and 
 e) exporting the selected given batch. 
 
     
     
       3. The method of  claim 1 , wherein the selecting the reference batch comprises one or more of the steps of:
 a) further selecting a plurality of reference batches from the screened batch data, and for the plurality of further selected reference batches, calculating quantitative statistical measures for each batch of the plurality as compared with average values of the plurality; 
 b) displaying one or more of the plurality of further selected reference batches together to a user with variable profiles in a given view, along with a timeline that represents progression in time of each batch of the plurality 
 c) enabling user to select a subset of batches of the screened batch data for the batch alignment based on the given view and domain knowledge of the user; and 
 d) enabling the user to discard at least one of the selected batch variables to join the batch alignment. 
 
     
     
       4. The method of  claim 1 , wherein the selecting the one or more batch variables comprises one or more of the steps of:
 a) discarding at least one of the one or more batch variables having flat trajectories or trajectories inconsistent with each other; 
 b) selecting a subset of batches from the one or more batch variables; 
 c) grouping correlated variables of the one or more batch variables; 
 d) identifying trajectory shapes of the one or more batch variables for a given phase associated with each of the one or more batch variables; 
 e) calculating a smoothness index for at least one of the one or more batch variables and the given phase; 
 f) calculating a curvature index for the at least one of the one or more batch variables and the given phase; 
 g) calculating a consistency index for the at least one of the one or more batch variables and the given phase; 
 h) determining an alignment score associated with at least one of the one or more batch variables and the given phase; 
 i) displaying one or more of the following to the user: the discarded batch variables, the selected subset of batches, the grouped correlated variables, the identified trajectory shapes, the calculated smoothness index, the calculated curvature index, the calculated consistency index, the determined alignment score; and 
 j) providing the user with one or more suggestions to further select any of the at least one of the one or more batch variables and the given phase. 
 
     
     
       5. The method of  claim 1 , wherein the estimating the one or more weightings comprises one or more of the steps of:
 a) pre-calculating weighting coefficients of one or more default variables of the selected batch variables according to a trajectory shape associated with the one or more default variables; 
 b) adjusting at least one of the one or more weightings based on a rank and standard deviation of at least one of the selected batch variables; 
 c) multiplying the one or more weightings with a corresponding consistency index; and 
 d) further adjusting the one or more weightings in an iterative manner. 
 
     
     
       6. The method of  claim 1 , wherein the performing the batch alignment in the online mode further comprises one or more of the steps of:
 a) determining a phase of a current batch of the subject industrial batch process for alignment; 
 b) selecting the one or more estimated variable weightings associated with the current batch phase; 
 c) adjusting the one or more estimated variable weightings associated with dynamic time warping (DTW) based on information from a previous sequential alignment point of the subject industrial batch process; 
 d) estimating a current batch maturity of the subject industrial batch process based on the dynamic time warping (DTW); 
 e) adjusting the one or more estimated variable weightings based on the current batch maturity estimation from the dynamic time warping (DTW); 
 f) adjusting one or more the dynamic time warping (DTW) batch maturity bounds based on the current batch maturity estimation; 
 g) modifying a starting point of the sliding window associated with the subject industrial batch process; 
 h) calculating a prediction and detecting a change in trajectory shapes of the selected batch variables; 
 i) increasing a size of the sliding window based upon a proximity of a lower bound of the dynamic time warping (DTW) solution to a starting point of the sliding window; 
 j) identifying one or more alignment times; 
 k) checking the size of the sliding window for an increase based upon the identified alignment times; 
 l) further adjusting the one or more batch maturity bounds based on the current alignment window size; 
 m) performing the alignment of the current batch; and 
 n) repeating steps (a) to (m) over time for the current batch. 
 
     
     
       7. The method of  claim 6 , further comprising one or more of the steps of:
 a) defining the batch alignment as a free-end-point problem with a cumulative distance; 
 b) generating a grid based upon the defined batch alignment, a number of data points of the current batch, and a number of data points of the reference batch; and 
 c) determining a warping path associated with the generated grid by traversing the generated grid in a monotonic fashion. 
 
     
     
       8. The method of  claim 1 , wherein performing the batch alignment includes performing alignment for a sliding window associated with the subject industrial batch is further based upon a modified Golden Section Search (GSS). 
     
     
       9. The method of  claim 4 , wherein the grouping the correlated variables further comprises one or more of the steps of:
 a) building a principal component analysis (PCA) model based on unfolding of the screened batch data; and 
 b) applying K-means clustering to one or more scores of the principal component analysis (PCA) model. 
 
     
     
       10. A computer-implemented system for performing automated batch data alignment for modeling, monitoring, and controlling an industrial batch process, the system comprising:
 a processor; and 
 a memory with computer code instructions stored thereon, the memory operatively coupled to the processor such that, when executed by the processor, the computer code instructions cause the computer system to implement: 
 a modeler engine configured to: 
 load batch data from a plant historian database for a subject industrial batch process; 
 scale the loaded batch data for batch alignment; 
 screen and remove outliers of batch data from the scaled batch data; 
 select a reference batch as basis of the batch alignment from the screened batch data; 
 define and add or modify one or more batch phases associated with the screened batch data for the batch alignment; 
 select one or more batch variables associated with the screened batch data based on at least one of: (i) one or more profiles of the one or more batch variables and (ii) one or more curvatures of the one or more batch variables; 
 estimate one or more variable weightings based upon the selected batch variables associated with the batch alignment; 
 select a given batch of the screened batch data to perform the batch alignment against the reference batch; 
 adjust one or more tuning parameters associated with the batch alignment; and 
 perform the batch alignment in at least one of an offline mode and an online mode, wherein performing the batch alignment includes performing alignment for a sliding window associated with the subject industrial batch process based upon a continuous objective function that includes a linearly interpolated value solved using dynamic time interpolation (DTI). 
 
     
     
       11. The system of  claim 10 , the modeler engine being further configured to screen and remove the outliers of the batch data including at least one of:
 a) further screening measurements of the selected batch variables for irregular behaviors as compared to behaviors associated with other batches of the screened batch data; 
 b) repairing data of the screened batch data associated with the irregular behaviors; 
 c) removing one or more invalid batches from the screened batch data after the data repair including batches that include different variation profiles as compared to the other batches over a time interval; 
 d) resampling the selected batch variables with a base sampling rate; and 
 e) exporting the selected given batch. 
 
     
     
       12. The system of  claim 10 , the modeler engine being further configured to select the reference batch including one or more of:
 a) further selecting a plurality of reference batches from the screened batch data, and for the plurality of further selected reference batches, calculating quantitative statistical measures for each batch of the plurality as compared with average values of the plurality; 
 b) displaying one or more of the plurality of further selected reference batches together to a user with variable profiles in a given view, along with a timeline that represents progression in time of each batch of the plurality 
 c) enabling user to select a subset of batches of the screened batch data for the batch alignment based on the given view and domain knowledge of the user; and 
 d) enabling the user to discard at least one of the selected batch variables to join the batch alignment. 
 
     
     
       13. The system of  claim 10 , the modeler engine further configured to select the one or more batch variables including one or more of:
 a) discarding at least one of the one or more batch variables having flat trajectories or trajectories inconsistent with each other; 
 b) selecting a subset of batches from the one or more batch variables; 
 c) grouping correlated variables of the one or more batch variables; 
 d) identifying trajectory shapes of the one or more batch variables for a given phase associated with each of the one or more batch variables; 
 e) calculating a smoothness index for at least one of the one or more batch variables and the given phase; 
 f) calculating a curvature index for the at least one of the one or more batch variables and the given phase; 
 g) calculating a consistency index for the at least one of the one or more batch variables and the given phase; 
 h) determining an alignment score associated with at least one of the one or more batch variables and the given phase; 
 i) displaying one or more of the following to the user: the discarded batch variables, the selected subset of batches, the grouped correlated variables, the identified trajectory shapes, the calculated smoothness index, the calculated curvature index, the calculated consistency index, the determined alignment score; and 
 j) providing the user with one or more suggestions to further select any of the at least one of the one or more batch variables and the given phase. 
 
     
     
       14. The system of  claim 10 , the modeler engine further configured to estimate the one or more weightings including one or more of:
 a) pre-calculating weighting coefficients of one or more default variables of the selected batch variables according to a trajectory shape associated with the one or more default variables; 
 b) adjusting at least one of the one or more weightings based on a rank and standard deviation of at least one of the selected batch variables; 
 c) multiplying the one or more weightings with a corresponding consistency index; and 
 d) further adjusting the one or more weightings in an iterative manner. 
 
     
     
       15. The system of  claim 10 , the modeler engine further configured to perform the batch alignment in the online mode including at least one of:
 a) determining a phase of a current batch of the subject industrial batch process for alignment; 
 b) selecting the one or more estimated variable weightings associated with the current batch phase; 
 c) adjusting the one or more estimated variable weightings associated with dynamic time warping (DTW) based on information from a previous sequential alignment point of the subject industrial batch process; 
 d) estimating a current batch maturity of the subject industrial batch process based on the dynamic time warping (DTW); 
 e) adjusting the one or more estimated variable weightings based on the current batch maturity estimation from the dynamic time warping (DTW); 
 f) adjusting one or more the dynamic time warping (DTW) batch maturity bounds based on the current batch maturity estimation; 
 g) modifying a starting point of the sliding window associated with the subject industrial batch process; 
 h) calculating a prediction and detecting a change in trajectory shapes of the selected batch variables; 
 i) increasing a size of the sliding window based upon a proximity of a lower bound of the dynamic time warping (DTW) solution to a starting point of the sliding window; 
 j) identifying one or more alignment times; 
 k) checking the size of the sliding window for an increase based upon the identified alignment times; 
 l) further adjusting the one or more batch maturity bounds based on the current alignment window size; 
 m) performing the alignment of the current batch; and 
 n) repeating steps (a) to (m) over time for the current batch. 
 
     
     
       16. The system of  claim 15 , the modeler engine being further configured to perform at least one of:
 a) defining the batch alignment as a free-end-point problem with a cumulative distance; 
 b) generating a grid based upon the defined batch alignment, a number of data points of the current batch, and a number of data points of the reference batch; and 
 c) determining a warping path associated with the generated grid by traversing the generated grid in a monotonic fashion. 
 
     
     
       17. The system of  claim 10 , wherein performing the batch alignment includes performing alignment for a sliding window associated with the subject industrial batch is further based upon a modified Golden Section Search (GSS). 
     
     
       18. The system of  claim 13 , the modeler engine being further configured to group the correlated variables including at least one of:
 a) building a principal component analysis (PCA) model based on unfolding of the screened batch data; and 
 b) applying K-means clustering to one or more scores of the principal component analysis (PCA) model. 
 
     
     
       19. A computer program product comprising:
 a non-transitory computer-readable storage medium having code instructions stored thereon, the storage medium operatively coupled to a processor, such that, when executed by the processor for modeling, monitoring, and controlling an industrial batch process, the computer code instructions cause the processor to: 
 load batch data from a plant historian database for a subject industrial batch process; 
 scale the loaded batch data for batch alignment; 
 screen and remove outliers of batch data from the scaled batch data; 
 select a reference batch as basis of the batch alignment from the screened batch data; 
 define and add or modify one or more batch phases associated with the screened batch data for the batch alignment; 
 select one or more batch variables associated with the screened batch data based on at least one of: (i) one or more profiles of the one or more batch variables and (ii) one or more curvatures of the one or more batch variables; 
 estimate one or more variable weightings based upon the selected batch variables associated with the batch alignment; 
 select a given batch of the screened batch data to perform the batch alignment against the reference batch; 
 adjust one or more tuning parameters associated with the batch alignment; and 
 perform the batch alignment in at least one of an offline mode and an online mode, wherein performing the batch alignment includes performing alignment for a sliding window associated with the subject industrial batch process based upon a continuous objective function that includes a linearly interpolated value solved using dynamic time interpolation (DTI).